tszzl: Most new code will be written by AI models, so fewer security holes
tszzl · x · 2026-10-22
In an ongoing debate on AI and cybersecurity, tszzl pushes back on the fear that AI-generated code will be riddled with vulnerabilities: he expects most new code will be written by models like fable/astra and therefore won't be as security-broken as human-written code. A notable counterpoint to the 'AI worsens security' narrative.
More from Safety
- Dario says government should gate AI access; critics say Anthropic's bill doesn't do that — AlexTensor · 2026-10-22
- One man's AI agents breached 100+ company sites at ~$50 in tokens per hack — srimisra · 2026-10-22
- When the optimizer controls its own scoreboard: the hidden runaway path of recursive self-improvement — Gauri_the_great · 2026-10-22
- The world can't agree on AI safety — so why does the EU keep trying? — kindermaxi123 · 2026-10-22
- AI safety researcher urges AI lab employees to stop escalating model capabilities amid cyberattacks — max_paperclips · 2026-10-22
- Researcher calls for national-scale AI cybersecurity push: systems will finally be secure once backlog is paid down — max_paperclips · 2026-10-22